Personalized Text to Speech Synthesis through Few Shot Speaker Adaptation with Contrastive Learning
Bibliographic record
Abstract
Personalized text-to-speech (TTS) synthesis has the goal of producing natural and expressive speech that emulates the voice of a target speaker with a minimum of data. The models of traditional neural TTS, including Tacotron 2 and Fast Speech 2, need to be trained in large amounts of speaker-specific data and can thus not easily be personalized quickly. We suggest CL-FS-TTS (Contrastive Learning based Few-Shot Text-to-Speech) to solve this problem, a new framework that uses contrastive speaker representation learning to adapt the speaker using only 1030 seconds of reference audio. The CL-FS-TTS architecture has two encoders: a content encoder that identifies linguistic features of text and a speaker encoder trained with the help of supervised contrastive learning to maximize speaker dissimilarity. In adaptation, the model matches speaker embeddings with generated mel-spectrograms with a contrastive consistency loss, enhancing voice and prosodic consistency. We compare CL-FS-TTS with Tacotron 2, Fast Speech 2, AdaSpeech, YourTTS, and Meta-TTS in terms of Mean Opinion Score (MOS), Speaker Similarity Score (SSS), Mel Cepstral Distortion (MCD) and Word Error Rate (WER). The experimental outcomes indicate that CL-FS-TTS has a higher naturalness and similarity of the speaker besides 40% less adaptation time in comparison with baselines. The suggested model lays the foundation of an efficient and strong model of high-quality personalized TTS synthesis in the situation of data scarcity.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".